The repository bojieli/ai-agent-book provides an open‑source, multilingual textbook that structures AI agents around the core formula Agent = LLM + context + tools. Spanning ten chapters, it progresses from foundational concepts (prompt engineering, KV cache, user memory, RAG) to advanced topics such as tool discovery via the MCP protocol, coding‑agent construction, asynchronous and multimodal interaction, evaluation frameworks, post‑training (SFT/RL), continual evolution through trajectory‑based learning, and multi‑agent collaboration. Each chapter is accompanied by a set of hands‑on experiments; the book reports 109 experiments in total, with chapter‑specific counts ranging from 4 to 19 (e.g., chapter 5 offers 16 coding‑agent experiments, chapter 8 includes 19 post‑training exercises). The experiments rely on external repositories cloned at fixed commits (e.g., anthropics/claude‑quickstarts, browser‑use/browser‑use, google‑research/android_world, huggingface/GAIA, xlang‑ai/OSWorld, SWE‑bench, bojieli/minimind, etc.) to reproduce environments for computer use, robotics, benchmarking, and model training. Technical execution assumes Python 3.11‑3.13, with dependency management via uv or pip, and optional PDF/EPUB generation requiring pandoc, xelatex, and the ElegantBook LaTeX class. All source, translations (15 languages), and build scripts are hosted under an Apache 2.0 license.